NON-INTRUSIVE EXHAUST GAS SENSOR MONITORING
A method for monitoring an exhaust gas sensor coupled in an engine exhaust is provided. In one embodiment, the method comprises indicating exhaust gas sensor degradation based on characteristics of a distribution of extreme values of a plurality of sets of lambda differentials collected during selected operating conditions. In this way, the exhaust gas sensor may be monitored in a non-intrusive manner.
Latest Ford Patents:
The present disclosure relates to an exhaust gas sensor in a motor vehicle.BACKGROUND AND SUMMARY
An exhaust gas sensor may be positioned in an exhaust system of a vehicle to detect an air/fuel ratio of exhaust gas exhausted from an internal combustion engine of the vehicle. The exhaust gas sensor readings may be used to control operation of the internal combustion engine to propel the vehicle.
Degradation of an exhaust gas sensor may cause engine control degradation that may result in increased emissions and/or reduced vehicle drivability. Accordingly, accurate determination of exhaust gas sensor degradation may reduce the likelihood of engine control based on readings from a degraded exhaust gas senor. In particular, an exhaust gas sensor may exhibit six discrete types of degradation behavior. The degradation behavior types may be categorized as asymmetric type degradation (e.g., rich-to-lean asymmetric delay, lean-to-rich asymmetric delay, rich-to-lean asymmetric slow response, lean-to-rich asymmetric slow response) that affects only lean-to-rich or rich-to-lean exhaust gas sensor response rates, or symmetric type degradation (e.g., symmetric delay, symmetric slow response) that affects both lean-to-rich and rich-to-lean exhaust gas sensor response rates. The delay type degradation behaviors may be associated with the initial reaction of the exhaust gas sensor to a change in exhaust gas composition and the slow response type degradation behaviors may be associated with a duration after an initial exhaust gas sensor response to transition from a rich-to-lean or lean-to-rich exhaust gas sensor output.
Previous approaches to monitoring exhaust gas sensor degradation, particularly identifying one or more of the six degradation behaviors, have relied on intrusive data collection. That is, an engine may be purposely operated with one or more rich to lean or lean to rich transitions to monitor exhaust gas sensor response. However, these excursions may be restricted to particular operating conditions that do not occur frequently enough to accurately monitor the sensor, such as during deceleration fuel shut off conditions. Further, these excursions may increase engine operation at non-desired air/fuel ratios that result in increased fuel consumption and/or increased emissions.
The inventors herein have recognized the above issues and identified a non-intrusive approach for determining exhaust gas sensor degradation. In one embodiment, a method of monitoring an exhaust gas sensor coupled in an engine exhaust comprises indicating exhaust gas sensor degradation based on characteristics of a distribution of extreme values of a plurality of sets of lambda differentials collected during selected operating conditions.
In this way, exhaust gas sensor degradation may be indicated by monitoring characteristics of a distribution of extreme values from multiple sets of successive lambda samples in steady state operating conditions. In one example, the characteristics may be a mode and central peak of a generalized extreme value (GEV) distribution of the extreme lambda differentials collected during steady state operating conditions. Asymmetric delay or asymmetric slow response degradation may be determined based on the magnitude of the central peak and/or the magnitude of the mode. Further classification, for example symmetric delay or symmetric slow response, may be based on a determined sensor delay or a determined sensor time constant.
By determining degradation of an exhaust gas sensor using a non-intrusive approach with data collected during selected operating conditions, exhaust gas sensor degradation monitoring may be performed in a simple manner. Further, by using the exhaust gas sensor output to determine which of the seven degradation behaviors the sensor exhibits, closed loop feedback control may be improved by tailoring engine control (e.g., fuel injection amount and/or timing) responsive to indication of the particular degradation behavior of the exhaust gas sensor to reduce the impact on vehicle drivability and/or emissions due to exhaust gas sensor degradation.
The above advantages and other advantages, and features of the present description will be readily apparent from the following Detailed Description when taken alone or in connection with the accompanying drawings.
It should be understood that the summary above is provided to introduce in simplified form a selection of concepts that are further described in the detailed description. It is not meant to identify key or essential features of the claimed subject matter, the scope of which is defined uniquely by the claims that follow the detailed description. Furthermore, the claimed subject matter is not limited to implementations that solve any disadvantages noted above or in any part of this disclosure.
The following description relates to an approach for determining degradation of an exhaust gas sensor. More particularly, the systems and methods described below may be implemented to determine exhaust gas sensor degradation based on recognition of any one of six discrete types of behavior associated with exhaust gas sensor degradation.
Combustion chamber 30 may receive intake air from intake manifold 44 via intake passage 42 and may exhaust combustion gases via exhaust passage 48. Intake manifold 44 and exhaust passage 48 can selectively communicate with combustion chamber 30 via respective intake valve 52 and exhaust valve 54. In some embodiments, combustion chamber 30 may include two or more intake valves and/or two or more exhaust valves.
In this example, intake valve 52 and exhaust valves 54 may be controlled by cam actuation via respective cam actuation systems 51 and 53. Cam actuation systems 51 and 53 may each include one or more cams and may utilize one or more of cam profile switching (CPS), variable cam timing (VCT), variable valve timing (VVT) and/or variable valve lift (VVL) systems that may be operated by controller 12 to vary valve operation. The position of intake valve 52 and exhaust valve 54 may be determined by position sensors 55 and 57, respectively. In alternative embodiments, intake valve 52 and/or exhaust valve 54 may be controlled by electric valve actuation. For example, cylinder 30 may alternatively include an intake valve controlled via electric valve actuation and an exhaust valve controlled via cam actuation including CPS and/or VCT systems.
Fuel injector 66 is shown arranged in intake passage 44 in a configuration that provides what is known as port injection of fuel into the intake port upstream of combustion chamber 30. Fuel injector 66 may inject fuel in proportion to the pulse width of signal FPW received from controller 12 via electronic driver 68. Fuel may be delivered to fuel injector 66 by a fuel system (not shown) including a fuel tank, a fuel pump, and a fuel rail. In some embodiments, combustion chamber 30 may alternatively or additionally include a fuel injector coupled directly to combustion chamber 30 for injecting fuel directly therein, in a manner known as direct injection.
Ignition system 88 can provide an ignition spark to combustion chamber 30 via spark plug 92 in response to spark advance signal SA from controller 12, under select operating modes. Though spark ignition components are shown, in some embodiments, combustion chamber 30 or one or more other combustion chambers of engine 10 may be operated in a compression ignition mode, with or without an ignition spark.
Exhaust gas sensor 126 is shown coupled to exhaust passage 48 of exhaust system 50 upstream of emission control device 70. Sensor 126 may be any suitable sensor for providing an indication of exhaust gas air/fuel ratio such as a linear oxygen sensor or UEGO (universal or wide-range exhaust gas oxygen), a two-state oxygen sensor or EGO, a HEGO (heated EGO), a NOx, HC, or CO sensor. In some embodiments, exhaust gas sensor 126 may be a first one of a plurality of exhaust gas sensors positioned in the exhaust system. For example, additional exhaust gas sensors may be positioned downstream of emission control 70.
Emission control device 70 is shown arranged along exhaust passage 48 downstream of exhaust gas sensor 126. Device 70 may be a three way catalyst (TWC), NOx trap, various other emission control devices, or combinations thereof. In some embodiments, emission control device 70 may be a first one of a plurality of emission control devices positioned in the exhaust system. In some embodiments, during operation of engine 10, emission control device 70 may be periodically reset by operating at least one cylinder of the engine within a particular air/fuel ratio.
Controller 12 is shown in
Furthermore, at least some of the above described signals may used in the exhaust gas sensor degradation determination method described in further detail below. For example, the inverse of the engine speed may be used to determine delays associated with the injection—intake—compression—expansion—exhaust cycle. As another example, the inverse of the velocity (or the inverse of the MAF signal) may be used to determine a delay associated with travel of the exhaust gas from the exhaust valve 54 to exhaust gas sensor 126. The above described examples along with other use of engine sensor signals may be used to determine the time delay between a change in the commanded air fuel ratio and the exhaust gas sensor response rate.
In some embodiments, exhaust gas sensor degradation determination may be performed in a dedicated controller 140. Dedicated controller 140 may include processing resources 142 to handle signal-processing associated with production, calibration, and validation of the degradation determination of exhaust gas sensor 126. In particular, a sample buffer (e.g., generating approximately 100 samples per second per engine bank) utilized to record the response rate of the exhaust gas sensor may be too large for the processing resources of a powertrain control module (PCM) of the vehicle. Accordingly, dedicated controller 140 may be operatively coupled with controller 12 to perform the exhaust gas sensor degradation determination. Note that dedicated controller 140 may receive engine parameter signals from controller 12 and may send engine control signals and degradation determination information among other communications to controller 12.
Note storage medium read-only memory 106 and/or processing resources 142 can be programmed with computer readable data representing instructions executable by processor 102 and/or dedicated controller 140 for performing the methods described below as well as other variants.
As discussed above, exhaust gas sensor degradation may be determined based on any one, or in some examples each, of six discrete behaviors indicated by delays in the response rate of air/fuel ratio readings generated by an exhaust gas sensor during rich-to-lean transitions and/or lean-to-rich transitions.
It will be appreciated that a degraded exhaust gas sensor may exhibit a combination of two or more of the above described degradation behaviors. For example, a degraded exhaust gas sensor may exhibit an asymmetric rich-to-lean response degradation behavior (i.e.,
Turning now to
Referring specifically to
If it is determined at 804 that the engine is not operating in steady-state conditions, method 800 returns to 802 to continue to determine engine operating parameters. If steady state conditions are determined, method 800 proceeds to 806 to calculate air/fuel ratio, or lambda, differentials for a given duration based on readings from the exhaust gas sensor being monitored (e.g., sensor 126). Lambda may be determined for a given number of samples over a given time duration, for example samples may be collected at a rate of 1 sample/96 ms for 60 seconds. For each sample, the difference between that determined lambda and the previous lambda may be calculated and stored in the memory of the controller.
The lambda differentials are plotted in a non-normal distribution, and then the characteristics of the distribution are determined at 808. In one example, all calculated lambda differentials may be plotted together, and a distribution curve may be drawn based on the plotted data. In another example, N data sets may be generated from the calculated differentials, and the extreme values from each data set of lambda differentials may be determined. For example, 100 data sets may be generated, and the highest and/or lowest value from each data set may be chosen as the extreme values. These extreme values may be plotted and a distribution curve determined. In one embodiment, determining a distribution curve based on extreme values may include a generalized extreme value (GEV) distribution:
Where k is the shape, σ is the scale, and μ is the location of the distribution curve.
In one embodiment, the characteristics of the distribution may include the magnitude of a central peak and of a mode of the distribution. The mode is the value that occurs most frequently in the distribution, and the central peak is the percentage of the data samples that have that value. In a GEV distribution, the mode may be determined by the equation:
A sensor degradation behavior may be determined based on the characteristics of the distribution at 810. For example, as will be explained in more detail below, the magnitude of the central peak (which indicates the degree of variation of the extreme lambda differentials) may indicate whether or not an asymmetric delay degradation behavior is present, as sensors with asymmetric delay type degradation may exhibit less variation than sensors without asymmetric delay. Additionally, the magnitude of the mode (which indicates whether the sensor output is biased rich or lean) may indicate whether or not an asymmetrical response degradation behavior is present. By determining the magnitude of the central peak and the magnitude of the mode, as well as determining other sensor parameters as will be described in more detail below, the sensor can be classified into one or more of the six discrete degradation behaviors, or be classified as not degraded. Determining the sensor degradation behavior based on the characteristics of the distribution will be described in more detail with regard to
At 812, method 800 comprises determining if sensor degradation is indicated. If no degradation is indicated (e.g., the characteristics of the distribution indicate that no degradation behavior is present), method 800 returns to continue to monitor the sensor. If degradation is indicated, method 800 proceeds to 814 to determine the whether the sensor degradation behavior exceeds a maximum value. As described above, sensor degradation may be indicated based on the characteristics of an extreme value distribution of lambda differentials. The characteristic that indicates degradation (e.g., the central peak or mode) may be analyzed to determine the extent of the degradation. For example, a central peak magnitude above a given first threshold may indicate an asymmetric delay degradation behavior. If the magnitude is above the first threshold by a sufficient amount, for example if it is 20% or more greater than the first threshold, the degradation behavior may exceed the maximum limit. If the degradation behavior exceeds the maximum value, this may indicate the sensor is damaged or otherwise non-functional and as such method 800 proceeds to 816 to notify an operator of the vehicle of the sensor degradation, for example by activating a malfunction indication light. If the degradation behavior does not exceed the maximum value, it may indicate that the sensor is still functional. However, to ensure adequate engine control to maintain engine emissions and fuel economy at a desired level, one or more engine operating parameters may be adjusted at 818, if desired. This may include adjusting fuel injection amount and/or timing, and may include adjusting control routines that are based on feedback from the degraded sensor to compensate for the identified degradation.
As explained above, method 800, as well as method 900 described with respect to
If the central peak is not less the first threshold, an asymmetric delay sensor degradation behavior is indicated. An example GEV distribution with a central peak greater than the first threshold is illustrated as curve 1102 of
Returning to 902, if the central peak is less than the first threshold, method 900 proceeds to 910 to determine if the mode of the distribution is less than a second threshold. As explained above, the mode is the lambda differential value that occurs in the distribution most frequently. A symmetric sensor, that is a sensor that does not display any asymmetric sensor degradation, will typically have a mode in a symmetric range centered around zero, bounded by a second and third threshold. The second and third thresholds can be determined in a manner similar to the first, central peak threshold. Example second and third thresholds, T2 and T3, are illustrated in
If the mode is smaller or larger than the symmetric range, asymmetric response type degradation behavior is indicated. If the mode is larger than the symmetric range, that is if the mode is not less than the second threshold, method 900 proceeds to 912 to indicate a rich-to-lean response degradation. In this case, the sensor experiences a delay in the response to a commanded rich to lean change, and thus spends less time at the commanded lean lambda, than at the commanded rich lambda. Thus, a greater amount of the lambda differentials will occur with values with a positive (lean) magnitude.
If the mode is less than the second threshold, method 900 proceeds to 914 to determine if the mode is greater than the third threshold. If not, the mode is therefore less than the symmetric range, and thus method 900 includes indicating a lean to rich response degradation at 916. An example GEV distribution curve with a mode less than the third threshold is illustrated as curve 1104 of
To determine which symmetric condition the sensor is exhibiting, method 900 includes determining if the sensor time delay is less than or equal to a nominal time delay at 918. The nominal sensor time delay is the expected delay in sensor response to a commanded air/fuel ratio change based on the delay from when the fuel is injected, combusted, and the exhaust travels from the combustion chamber to the exhaust sensor. The determined time delay may be when the sensor actually outputs a signal indicating the changed air/fuel ratio. If the time delay is not less than or equal to the nominal time delay, method 900 proceeds to 920 to indicate a symmetric delay.
If the time delay is less than or equal to the nominal time delay, method 900 proceeds to 922 to determine if a time constant of the sensor is less than or equal to a nominal time constant. The nominal time constant may be the time constant indicating how quickly the sensor responds to a commanded change in lambda, and may be determined off-line based on non-degraded sensor function. If the determined time constant is greater than the nominal time constant, it indicates a slow response rate, and thus at 924, if the time constant is not less than or equal to the nominal time constant, a symmetric response degradation behavior is indicated.
If the time constant is less than or equal to the nominal time constant, method 900 includes indicating no degradation at 926. No degradation is indicated due to the characteristics of the distribution indicating a symmetric behavior of the sensor, and both the sensor time constant and delay being similar to the nominal time constant and delay. Upon indicating a sensor behavior, whether one of the six discrete degradation behaviors or the no degradation behavior, method 900 exits.
Thus, the methods described with respect to
While the methods described with respect to
SVM is a supervised learning algorithm, where given a training set with known class information, a model is developed to classify the unknown test samples into different classes. The SVM processes a set of input data and predicts, for each given input, which of two possible classes the input is a member of, which makes the SVM a non-probabilistic binary linear classifier. In one embodiment, the SVM predicts whether or not the exhaust gas sensor is operating with one of the six sensor degradation behaviors. Typically, the SVM algorithm may be generated via a plurality of sets of training examples, each marked as belonging to one of two categories. The SVM training algorithm builds a model that assigns new examples into one category or the other. However, to differentiate among multiple classes, the SVM algorithm may be trained with multiple training sets, where each set is marked as belonging to one of the six classes of sensor degradation, or not belonging to that class.
An SVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap (sometimes referred to as a margin) that is as wide as possible. New examples are then mapped into that same space and predicted to belong to a category based on which side of the gap they fall on.
In the depicted embodiment, the SVM may classify unknown inputs into one of seven classes, the six sensor degradation behaviors described with respect to
SVMs use a signum function as a linear classifier to classify the unknown inputs into the two groups based on the training function wherein known inputs are used. Specifically, the known inputs are mapped onto high- or infinite-dimensional space and one or more hyper-planes are chosen that separate the inputs into the two spaced groups. In some embodiments, a hyper-plane that represents the largest margin of separation of the groups is chosen, while in other embodiments, a hyper-plane with a margin that allows for some degree of error in the inputs may be chosen, known as a slack margin. After the model is trained, unknown inputs can be entered and classified into one of the two groups. Typically, the output of the signum function is either +1 or −1, but either classification may be transformed into other values, e.g., −1 may be transformed to 0.
If the known inputs used to train the model cannot be separated using a linear classification, a transformation function may be used with a non-linear classification to separate the inputs. Example transformation functions include radial basis functions, linear transformations, etc. Additionally or alternatively, soft margins may be used to introduce some slack variables to the classification to allow some misclassification for outlier data points.
For sensor diagnostics, various input parameters into the SVM may be used. In one embodiment, the input parameters may include air amount (AM) such as mass airflow rate from MAF sensor, sensor temperature estimated based on engine operating conditions such as speed, load, etc., upstream exhaust gas sensor output (e.g., UEGO output), and downstream exhaust gas sensor output (e.g., HEGO output). In some embodiments, all the example inputs listed above may be used in the SVM. In other embodiments, only a subset of the input parameters may be used, such as sensor temperature and output.
Because there are seven possible sensor degradation behaviors (the six discrete degradation behaviors and no degradation), a multiple classification scheme is used to differentiate among the degradation behaviors. As will be described below, each input may be fed into the SVM algorithm seven times, and marked as belonging or not belonging to each of the seven degradation behaviors. Then each classification may be given a confidence level based on how close that input value is to the support vectors (e.g., the values that define the margin). For example, inputs that are further away from the support vectors may be given higher confidence levels. After all the samples in a duration have been classified for all the degradation behaviors, the behaviors may be ranked based on the confidence levels.
Method 1000 comprises determining at 1002 if the engine is running and if selected conditions are met. The selected conditions may include that the input parameters are operational, for example, that the UEGO and HEGO sensors are at a temperature whereby they are outputting functional readings. Further, the selected conditions may include that combustion is occurring in the cylinders of the engine, e.g. that the engine is not in a shut-down mode such as deceleration fuel shut-off (DFSO), or that the engine is operating in steady state conditions.
If it is determined that the engine is not running and/or the selected conditions are not met, method 1000 returns and does not monitor sensor function. If the engine is running and the selected conditions are met, method 1000 proceeds to 1004 to scale the input parameters and transform using a kernel function. As explained above, non-linear inputs may be transformed using the kernel function. Various kernel functions may be used, such as a linear, polynomial, radial basis function, sigmoid, and others. In one example, the radial basis function may be used:
In one embodiment, γ=600, and the soft basis parameter C=650, where xi, xj, . . . etc. are the input parameters (e.g., MAF, temperature, sensor output).
At 1006, the signum function output is calculated for each of the seven degradation behaviors. The signum function determines a sign, or classification, for the output of the SVM. The SMV may be trained based on known inputs, and may include a set of model parameters that are used to predict an output from unclassified input parameters. Each input parameter determined at a single instance of time may be fed into the SVM model, and a signum function output produced based on:
yi=+1 if (w,xi+b)≧1
yi=−1 if (w,xi+b)≦1
With a hyper-plane of w,xi+b=0, and where yi is the predicted class for the test input xi. w is defined by the trained SVM model based on the support vectors computed from optimizing the margin of the hyper-plane.
In one embodiment, to reduce the number of support vectors or to reduce the data size of the trained model for implementation in a vehicle, clustering can be used. Clustering includes an un-supervised learning where the data set is divided into different clusters or groups, so as to minimize the total distance of each point from the respective centroid. In a simpler language, the datapoints which are closer to each other are assigned to one cluster. This technique is employed to initially divide the training set into K (pre-defined number) clusters in each class and then the SVM algorithm is used where the original dataset is replaced by the centroids of each cluster. It was observed that the significant reduction in the number of support vectors could be achieved without loss in accuracy.
At 1008, the output of the signum function is given a confidence level. In one example, this may include determining the difference between the input value and the nearest support vector. A running total of the confidence levels for each output of each class of sensor degradation may be stored.
At 1010, the output of the signum function is added to the total of all previously calculated outputs and the count (C) is increased by 1. In doing so, over a given duration (j), which may start following an engine start and once the input parameters can be reliably sensed the method determines for each sample instance (i) a classification CL based on the calibrated and trained support vector machine.
Where CL is the output of the signum function, and is either set to 1 or −1, but with −1 converted to zero. Then, the routine adds CLi to the running count Cj:
The count (C) is compared to a first threshold at 1012, and if C is above the threshold (e.g., the duration is complete) method 1000 proceeds to 1016. If C is not above the threshold, method 1000 proceeds to 1014 to continue to the next sample.
At 1016, each degradation class is ranked based on the associated confidence levels determined at 1008, as well as other parameters, such as the frequency/distribution of classifications within the classes. At 1018, the at least highest ranked class is selected. The highest ranked class may be the degradation behavior that the sensor is most likely to be exhibiting. In some embodiments, more than one sensor degradation behavior may be present. In such cases, the top two classes may selected. The determination which top-ranked classes to select may be determined in a suitable manner dependent upon the stringency of the degradation determination.
At 1020, it is determined if sensor degradation is indicated (e.g., if a class other than no degradation is selected at 1018). If not, method 1000 proceeds to 1022 to maintain current operating parameters. If degradation is indicated, method 1000 proceeds to 1024 to indicate sensor degradation and take default action. The indication may include notification being sent to a driver via a message system, or may be the setting of a diagnostic code read by a diagnostic code reader in a service station, or various other indications such as a malfunction indicator lamp (MIL).
The duration (j) may a suitable duration. For example, in one embodiment, the duration may include idle operation, or operation under steady state conditions. The duration may be long enough to collect a suitable number of inputs in order to reliably test the sensor function, e.g., 10 seconds, 60 seconds, etc. Further, more than one duration of samples may be collected during operation of the sensor monitor. The plurality of durations may include combinations of the above durations. In some embodiments, the plurality of durations may occur successively, that is, without an engine shutdown occurring between them.
The indication of degradation is as determined by method 1000 may be based on the parameter readings occurring during the indicated duration. For example, the indication of degradation may be based on parameter readings occurring during idle operation as explained above. Further, the indication of degradation may be based on parameter readings occurring engine temperature is above light-off temperature, or may be based on parameter readings occurring when engine speed is constant, etc.
While the method described above with respect to
Thus, the method of
It will be appreciated that the configurations and methods disclosed herein are exemplary in nature, and that these specific embodiments are not to be considered in a limiting sense, because numerous variations are possible. For example, the above technology can be applied to V-6, I-4, I-6, V-12, opposed 4, and other engine types. The subject matter of the present disclosure includes all novel and non-obvious combinations and sub-combinations of the various systems and configurations, and other features, functions, and/or properties disclosed herein.
The following claims particularly point out certain combinations and sub-combinations regarded as novel and non-obvious. These claims may refer to “an” element or “a first” element or the equivalent thereof. Such claims should be understood to include incorporation of one or more such elements, neither requiring nor excluding two or more such elements. Other combinations and sub-combinations of the disclosed features, functions, elements, and/or properties may be claimed through amendment of the present claims or through presentation of new claims in this or a related application. Such claims, whether broader, narrower, equal, or different in scope to the original claims, also are regarded as included within the subject matter of the present disclosure.
1. A method of monitoring an exhaust gas sensor coupled in an engine exhaust, comprising:
- indicating exhaust gas sensor degradation based on characteristics of a distribution of extreme values of a plurality of sets of lambda differentials collected during selected operating conditions.
2. The method of claim 1, wherein the distribution is a generalized extreme value (GEV) distribution, and wherein the characteristics include a magnitude of a mode and of a central peak of the GEV distribution.
3. The method of claim 2, wherein if the magnitude of the central peak is greater than a threshold, indicating an asymmetric delay sensor degradation.
4. The method of claim 3, wherein if an expected mean air/fuel ratio is greater than a determined mean air/fuel ratio at idle, indicating a rich to lean delay sensor degradation, and if the expected mean air/fuel ratio is less than the determined mean air/fuel ratio at idle, indicating a lean to rich delay sensor degradation.
5. The method of claim 3, wherein if the magnitude of the central peak is less than the threshold and the magnitude of the mode is outside a symmetric range, indicating an asymmetric response sensor degradation.
6. The method of claim 5, wherein if the magnitude of the mode is less than the symmetric range, indicating a lean to rich response sensor degradation, and if the magnitude of the mode is greater than the symmetric range, indicating a rich to lean response sensor degradation.
7. The method of claim 5, wherein if the magnitude of the mode is in the symmetric range, indicating no degradation or a symmetric sensor degradation.
8. The method of claim 7, further comprising indicating a symmetric delay sensor degradation if a determined time delay is greater than a nominal time delay, and indicating a symmetric response sensor degradation if a determined time constant is greater than a nominal time.
9. The method of claim 1, wherein the selected operating conditions further comprise steady state operating conditions.
10. The method of claim 1, further comprising adjusting a fuel injection amount and/or timing based on the indicated degradation.
11. A system for a vehicle, comprising:
- an engine including a fuel injection system;
- an exhaust gas sensor coupled in an exhaust system of the engine; and
- a controller including instructions executable to: indicate exhaust gas sensor degradation based on characteristics of a distribution of extreme values of a plurality of sets of lambda differentials collected during steady state operating conditions; and adjust an amount and/or timing of fuel injection based on the indicated sensor degradation.
12. The system of claim 11, wherein the instructions are further executable to notify an operator of the vehicle if the indicated sensor degradation exceeds a threshold.
13. The system of claim 11, wherein the distribution is a generalized extreme value (GEV) distribution, and wherein the characteristics include a magnitude of a mode and of a central peak of the GEV distribution.
14. A method of monitoring an oxygen sensor coupled in an engine exhaust, comprising:
- indicating an asymmetric delay sensor degradation if a first characteristic of a distribution of extreme values of a plurality of sets of lambda differentials exceeds a first threshold; and
- indicating an asymmetric response sensor degradation if the first characteristic is below the first threshold and a second characteristic of the distribution is outside a second threshold range.
15. The method of claim 14, wherein the first characteristic a magnitude of a central peak of the distribution and the second characteristic is a magnitude of a mode of the distribution.
16. The method of claim 14, further comprising indicating a no fault or symmetric sensor degradation if the first characteristic is below the first threshold and the second characteristic is within the second threshold range.
17. The method of claim 16, further comprising indicating a symmetric delay sensor degradation if a determined time delay of the sensor is greater than a nominal time delay, and indicating a symmetric response sensor degradation if a determined time constant of the sensor is greater than a nominal time constant.
18. The method of claim 14, further comprising adjusting a fuel injection amount based on an indicated sensor degradation.
19. The method of claim 14, wherein the lambda differentials are collected during steady state operating conditions.
20. A method of monitoring an exhaust gas sensor, comprising:
- applying a set of inputs for a given sample to a support vector machine to generate a classification output;
- recording a plurality of classification outputs for a plurality of successive samples over a duration;
- ranking each of the plurality of classifications based on confidence levels associated with each classification output; and
- indicating a sensor condition based on at least the highest ranked classification.
Filed: Jan 18, 2012
Publication Date: Jul 18, 2013
Patent Grant number: 9074513
Applicant: FORD GLOBAL TECHNOLOGIES, LLC (Dearborn, MI)
Inventors: Imad Hassan Makki (Dearborn Heights, MI), Pankaj Kumar (Houston, TX), James Michael Kerns (Trenton, MI), Hassene Jammoussi (Houston, TX)
Application Number: 13/352,885
International Classification: F02D 41/04 (20060101); G01N 33/00 (20060101);